Prediction of tapered steel plate girders shear strength using multigene genetic programming

Tapered steel plate girders have been widely used in the construction of bridges and heavy industry structures. However, predicting their structural behavior, particularly in shear, is challenging due to their non-prismatic sections and ill-understood influencing factors. Therefore, simplified analytical and classic regression techniques may not be able to capture the underlying nonlinear relationships controlling the shear behavior of tapered plate girders. In this study, multigene genetic programming (MGGP) was utilized to explore such complex relationships, and subsequently develop robust predictive expressions for the tapered steel plate girders shear strength. Attributed to the lack of a large experimental dataset, a nonlinear finite element model (FEM) was first developed and validated against available experimental results in literature. The FEM was subsequently employed to generate a matrix of 211 numerical results to augment 200 more FEM results compiled from previous studies, to cover a wider range of design parameters. The entire dataset was then used in the training and testing of the MGGP predictive expressions. The prediction accuracy of the developed expressions was evaluated against that of other existing expressions. The results showed that the adopted MGGP approach, guided be mechanics understanding, produced elegant predictive expressions with high level of accuracy and generalizability compared to other existing ones examined herein. As such, the developed expressions present an efficient prediction tool that can be adopted by design standards for estimating the ultimate shear strength of tapered steel girders. Finally, reliability analysis is performed on the developed expressions to introduce strength reduction factors in order to satisfy target design conservatism.